Feature Selection using Modified Charged System Search for Intrusion Detection Systems in Cloud Environment

DOI: NA

A novel feature selection technique based on a metaheuristic search algortihm is implemented for Intrusion Detection System. The proposed Modified Charged System Search algorithm selects optimal feature subset to give an efficient IDS with higher classification accuracy. The results are evaluated on dataset and presented in the paper.

Introduction

Usage

• Clone this repository
• Install the dependencies: pip install -r requirements.txt (use virtual environment)
• Run the CSS-FS.ipynb Jupyter notebook end-to-end for CSS Feature Selection
• Copy paste the selected features to input defined in classifiers.ipynb Jupyter notebook to evaluate using different classifiers

Dataset

The experiments are performed on NSL-KDD and 10% KDD Cup'99 Dataset. These dataset were pre-processed and normalized before use. It can be obtained from the following source.

NSL-KDD Dataset KDD Cup'99 Dataset

Results

The following results show the performance MCSS algortihm.

This figure shows variying classification accuracy for different number of features selected during an instance in search thus a need for feature selection.

This figure shows fast convergence of the MCSS algorithm

This figure shows postions of different agent during instances of search and its convergence towards the end. The few particles which do not converge are present as an improvement to give more exploration to the search.

Authors

• Shivam Shakti
• Partha Ghosh
• Santanu Phadikar

Copyright

This paper has been accepted and presented in SCESM 2017.